Allintra: guided adaptive quantization
In this CL, we introduce a new adaptive quantization scheme.
We read an external rate distribution map and use it to determine
quantizers for each superblock.
These following steps apply:
(1). Read rate distribution from file
(2). Estimate rate cost for each 16x16 block in preprocess stage
(3). Determine quantizer for each super block based on the relation
between two rate distributions.
This mode requires --deltaq-mode=3, since it makes use of its
information in the preprocess stage to determine quantizer.
We add two command line options (need both):
(1). --enable-rate-guide-deltaq=1. It turns on the new feature.
(2). --rate-distortion-info=input. It reads the rate distribution
from the input file.
This new scheme aims to improve perceptual quality. The quality of
the rate distribution map guides the adaptive quantization.
The perceptual quality of the final encoded image could be substantially
improved if the input rate distrbution map is well designed.
Change-Id: I44c424ba5221daa568dd072d16b3eb14adf2704b
diff --git a/aom/aomcx.h b/aom/aomcx.h
index 8b2722f..bd3f277 100644
--- a/aom/aomcx.h
+++ b/aom/aomcx.h
@@ -1484,6 +1484,18 @@
*/
AV1E_SET_QUANTIZER_ONE_PASS = 159,
+ /*!\brief Codec control to enable the rate distribution guided delta
+ * quantization in all intra mode. It requires --deltaq-mode=3, also
+ * an input file which contains rate distribution for each 16x16 block,
+ * passed in by --rate-distribution-info=rate_distribution.csv.
+ */
+ AV1E_ENABLE_RATE_GUIDE_DELTAQ = 160,
+
+ /*!\brief Codec control to set the input file for rate distribution used
+ * in all intra mode. It requires --enable-rate-guide-deltaq=1.
+ */
+ AV1E_SET_RATE_DISTRIBUTION_INFO = 161,
+
// Any new encoder control IDs should be added above.
// Maximum allowed encoder control ID is 229.
// No encoder control ID should be added below.
@@ -2060,6 +2072,12 @@
AOM_CTRL_USE_TYPE(AV1E_SET_PARTITION_INFO_PATH, const char *)
#define AOM_CTRL_AV1E_SET_PARTITION_INFO_PATH
+AOM_CTRL_USE_TYPE(AV1E_ENABLE_RATE_GUIDE_DELTAQ, unsigned int)
+#define AOM_CTRL_AV1E_ENABLE_RATE_GUIDE_DELTAQ
+
+AOM_CTRL_USE_TYPE(AV1E_SET_RATE_DISTRIBUTION_INFO, const char *)
+#define AOM_CTRL_AV1E_SET_RATE_DISTRIBUTION_INFO
+
AOM_CTRL_USE_TYPE(AV1E_SET_EXTERNAL_PARTITION, aom_ext_part_funcs_t *)
#define AOM_CTRL_AV1E_SET_EXTERNAL_PARTITION
diff --git a/apps/aomenc.c b/apps/aomenc.c
index 1cc45cd..dcab898 100644
--- a/apps/aomenc.c
+++ b/apps/aomenc.c
@@ -237,6 +237,8 @@
AV1E_SET_ENABLE_TX_SIZE_SEARCH,
AV1E_SET_LOOPFILTER_CONTROL,
AV1E_SET_AUTO_INTRA_TOOLS_OFF,
+ AV1E_ENABLE_RATE_GUIDE_DELTAQ,
+ AV1E_SET_RATE_DISTRIBUTION_INFO,
0 };
const arg_def_t *main_args[] = { &g_av1_codec_arg_defs.help,
@@ -437,6 +439,8 @@
#endif
&g_av1_codec_arg_defs.dv_cost_upd_freq,
&g_av1_codec_arg_defs.partition_info_path,
+ &g_av1_codec_arg_defs.enable_rate_guide_deltaq,
+ &g_av1_codec_arg_defs.rate_distribution_info,
&g_av1_codec_arg_defs.enable_directional_intra,
&g_av1_codec_arg_defs.enable_tx_size_search,
&g_av1_codec_arg_defs.loopfilter_control,
@@ -534,6 +538,8 @@
const char *vmaf_model_path;
#endif
const char *partition_info_path;
+ unsigned int enable_rate_guide_deltaq;
+ const char *rate_distribution_info;
aom_color_range_t color_range;
const char *two_pass_input;
const char *two_pass_output;
@@ -1131,6 +1137,12 @@
} else if (arg_match(&arg, &g_av1_codec_arg_defs.partition_info_path,
argi)) {
config->partition_info_path = arg.val;
+ } else if (arg_match(&arg, &g_av1_codec_arg_defs.enable_rate_guide_deltaq,
+ argi)) {
+ config->enable_rate_guide_deltaq = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &g_av1_codec_arg_defs.rate_distribution_info,
+ argi)) {
+ config->rate_distribution_info = arg.val;
} else if (arg_match(&arg, &g_av1_codec_arg_defs.use_fixed_qp_offsets,
argi)) {
config->cfg.use_fixed_qp_offsets = arg_parse_uint(&arg);
@@ -1541,6 +1553,16 @@
AV1E_SET_PARTITION_INFO_PATH,
stream->config.partition_info_path);
}
+ if (stream->config.enable_rate_guide_deltaq) {
+ AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
+ AV1E_ENABLE_RATE_GUIDE_DELTAQ,
+ stream->config.enable_rate_guide_deltaq);
+ }
+ if (stream->config.rate_distribution_info) {
+ AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
+ AV1E_SET_RATE_DISTRIBUTION_INFO,
+ stream->config.rate_distribution_info);
+ }
if (stream->config.film_grain_filename) {
AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder, AV1E_SET_FILM_GRAIN_TABLE,
diff --git a/av1/arg_defs.c b/av1/arg_defs.c
index a37958e..671567b 100644
--- a/av1/arg_defs.c
+++ b/av1/arg_defs.c
@@ -501,6 +501,16 @@
#endif
.partition_info_path = ARG_DEF(NULL, "partition-info-path", 1,
"Partition information read and write path"),
+ .enable_rate_guide_deltaq =
+ ARG_DEF(NULL, "enable-rate-guide-deltaq", 1,
+ "Enable rate guide deltaq (1), by default off (0)."
+ "It requires --deltaq-mode=3."
+ "If turned on, it requires to read an input file "
+ "by --rate-distribution-info"),
+ .rate_distribution_info =
+ ARG_DEF(NULL, "rate-distribution-info", 1,
+ "Rate distribution information input."
+ "It requires --enable-rate-guide-deltaq=1."),
.film_grain_test = ARG_DEF(
NULL, "film-grain-test", 1,
"Film grain test vectors (0: none (default), 1: test-1 2: test-2, "
diff --git a/av1/arg_defs.h b/av1/arg_defs.h
index 730b422..c7a640b 100644
--- a/av1/arg_defs.h
+++ b/av1/arg_defs.h
@@ -186,6 +186,8 @@
arg_def_t vmaf_model_path;
#endif
arg_def_t partition_info_path;
+ arg_def_t enable_rate_guide_deltaq;
+ arg_def_t rate_distribution_info;
arg_def_t film_grain_test;
arg_def_t film_grain_table;
#if CONFIG_DENOISE
diff --git a/av1/av1_cx_iface.c b/av1/av1_cx_iface.c
index aca0ab6..3e764dd 100644
--- a/av1/av1_cx_iface.c
+++ b/av1/av1_cx_iface.c
@@ -56,6 +56,8 @@
aom_tune_metric tuning;
const char *vmaf_model_path;
const char *partition_info_path;
+ unsigned int enable_rate_guide_deltaq;
+ const char *rate_distribution_info;
aom_dist_metric dist_metric;
unsigned int cq_level; // constrained quality level
unsigned int rc_max_intra_bitrate_pct;
@@ -235,6 +237,8 @@
AOM_TUNE_PSNR, // tuning
"/usr/local/share/model/vmaf_v0.6.1.json", // VMAF model path
".", // partition info path
+ 0, // enable rate guide deltaq
+ "./rate_map.csv", // rate distribution input
AOM_DIST_METRIC_PSNR, // dist_metric
10, // cq_level
0, // rc_max_intra_bitrate_pct
@@ -385,6 +389,8 @@
AOM_TUNE_PSNR, // tuning
"/usr/local/share/model/vmaf_v0.6.1.json", // VMAF model path
".", // partition info path
+ 0, // enable rate guide deltaq
+ "./rate_map.csv", // rate distribution input
AOM_DIST_METRIC_PSNR, // dist_metric
10, // cq_level
0, // rc_max_intra_bitrate_pct
@@ -715,6 +721,7 @@
AOM_SUPERBLOCK_SIZE_DYNAMIC);
RANGE_CHECK_HI(cfg, large_scale_tile, 1);
RANGE_CHECK_HI(extra_cfg, single_tile_decoding, 1);
+ RANGE_CHECK_HI(extra_cfg, enable_rate_guide_deltaq, 1);
RANGE_CHECK_HI(extra_cfg, row_mt, 1);
RANGE_CHECK_HI(extra_cfg, fp_mt, 1);
@@ -1465,6 +1472,9 @@
oxcf->partition_info_path = extra_cfg->partition_info_path;
+ oxcf->enable_rate_guide_deltaq = extra_cfg->enable_rate_guide_deltaq;
+ oxcf->rate_distribution_info = extra_cfg->rate_distribution_info;
+
oxcf->strict_level_conformance = extra_cfg->strict_level_conformance;
oxcf->kf_max_pyr_height = extra_cfg->kf_max_pyr_height;
@@ -2266,6 +2276,25 @@
return update_extra_cfg(ctx, &extra_cfg);
}
+static aom_codec_err_t ctrl_enable_rate_guide_deltaq(aom_codec_alg_priv_t *ctx,
+ va_list args) {
+ struct av1_extracfg extra_cfg = ctx->extra_cfg;
+ extra_cfg.enable_rate_guide_deltaq =
+ CAST(AV1E_ENABLE_RATE_GUIDE_DELTAQ, args);
+ return update_extra_cfg(ctx, &extra_cfg);
+}
+
+static aom_codec_err_t ctrl_set_rate_distribution_info(
+ aom_codec_alg_priv_t *ctx, va_list args) {
+ struct av1_extracfg extra_cfg = ctx->extra_cfg;
+ const char *str = CAST(AV1E_SET_RATE_DISTRIBUTION_INFO, args);
+ const aom_codec_err_t ret = allocate_and_set_string(
+ str, default_extra_cfg.rate_distribution_info,
+ &extra_cfg.rate_distribution_info, ctx->ppi->error.detail);
+ if (ret != AOM_CODEC_OK) return ret;
+ return update_extra_cfg(ctx, &extra_cfg);
+}
+
static aom_codec_err_t ctrl_set_film_grain_test_vector(
aom_codec_alg_priv_t *ctx, va_list args) {
struct av1_extracfg extra_cfg = ctx->extra_cfg;
@@ -2752,6 +2781,8 @@
&extra_cfg->second_pass_log);
check_and_free_string(default_extra_cfg.partition_info_path,
&extra_cfg->partition_info_path);
+ check_and_free_string(default_extra_cfg.rate_distribution_info,
+ &extra_cfg->rate_distribution_info);
check_and_free_string(default_extra_cfg.film_grain_table_filename,
&extra_cfg->film_grain_table_filename);
}
@@ -3713,6 +3744,17 @@
argv, err_string)) {
err = allocate_and_set_string(value, default_extra_cfg.partition_info_path,
&extra_cfg.partition_info_path, err_string);
+ } else if (arg_match_helper(&arg,
+ &g_av1_codec_arg_defs.enable_rate_guide_deltaq,
+ argv, err_string)) {
+ extra_cfg.enable_rate_guide_deltaq =
+ arg_parse_uint_helper(&arg, err_string);
+ } else if (arg_match_helper(&arg,
+ &g_av1_codec_arg_defs.rate_distribution_info,
+ argv, err_string)) {
+ err =
+ allocate_and_set_string(value, default_extra_cfg.rate_distribution_info,
+ &extra_cfg.rate_distribution_info, err_string);
} else if (arg_match_helper(&arg, &g_av1_codec_arg_defs.dist_metric, argv,
err_string)) {
extra_cfg.dist_metric = arg_parse_enum_helper(&arg, err_string);
@@ -4221,6 +4263,8 @@
{ AV1E_SET_SINGLE_TILE_DECODING, ctrl_set_single_tile_decoding },
{ AV1E_SET_VMAF_MODEL_PATH, ctrl_set_vmaf_model_path },
{ AV1E_SET_PARTITION_INFO_PATH, ctrl_set_partition_info_path },
+ { AV1E_ENABLE_RATE_GUIDE_DELTAQ, ctrl_enable_rate_guide_deltaq },
+ { AV1E_SET_RATE_DISTRIBUTION_INFO, ctrl_set_rate_distribution_info },
{ AV1E_SET_FILM_GRAIN_TEST_VECTOR, ctrl_set_film_grain_test_vector },
{ AV1E_SET_FILM_GRAIN_TABLE, ctrl_set_film_grain_table },
{ AV1E_SET_DENOISE_NOISE_LEVEL, ctrl_set_denoise_noise_level },
diff --git a/av1/encoder/allintra_vis.c b/av1/encoder/allintra_vis.c
index 0d935a7..0463856 100644
--- a/av1/encoder/allintra_vis.c
+++ b/av1/encoder/allintra_vis.c
@@ -35,11 +35,29 @@
// "compute_num_ai_workers()".
cpi->weber_bsize = BLOCK_8X8;
- if (cpi->mb_weber_stats) return;
+ if (cpi->oxcf.enable_rate_guide_deltaq) {
+ if (cpi->mb_weber_stats && cpi->prep_rate_estimates &&
+ cpi->ext_rate_distribution)
+ return;
+ } else {
+ if (cpi->mb_weber_stats) return;
+ }
CHECK_MEM_ERROR(cm, cpi->mb_weber_stats,
aom_calloc(cpi->frame_info.mi_rows * cpi->frame_info.mi_cols,
sizeof(*cpi->mb_weber_stats)));
+
+ if (cpi->oxcf.enable_rate_guide_deltaq) {
+ CHECK_MEM_ERROR(
+ cm, cpi->prep_rate_estimates,
+ aom_calloc(cpi->frame_info.mi_rows * cpi->frame_info.mi_cols,
+ sizeof(*cpi->prep_rate_estimates)));
+
+ CHECK_MEM_ERROR(
+ cm, cpi->ext_rate_distribution,
+ aom_calloc(cpi->frame_info.mi_rows * cpi->frame_info.mi_cols,
+ sizeof(*cpi->ext_rate_distribution)));
+ }
}
static int64_t get_satd(AV1_COMP *const cpi, BLOCK_SIZE bsize, int mi_row,
@@ -197,6 +215,20 @@
return sb_wiener_var;
}
+static int rate_estimator(const tran_low_t *qcoeff, int eob, TX_SIZE tx_size) {
+ const SCAN_ORDER *const scan_order = &av1_scan_orders[tx_size][DCT_DCT];
+
+ assert((1 << num_pels_log2_lookup[txsize_to_bsize[tx_size]]) >= eob);
+ int rate_cost = 1;
+
+ for (int idx = 0; idx < eob; ++idx) {
+ int abs_level = abs(qcoeff[scan_order->scan[idx]]);
+ rate_cost += (int)(log(abs_level + 1.0) / log(2.0)) + 1 + (abs_level > 0);
+ }
+
+ return (rate_cost << AV1_PROB_COST_SHIFT);
+}
+
void av1_calc_mb_wiener_var_row(AV1_COMP *const cpi, const int mi_row,
int16_t *src_diff, tran_low_t *coeff,
tran_low_t *qcoeff, tran_low_t *dqcoeff,
@@ -299,6 +331,13 @@
av1_quantize_fp_facade(coeff, pix_num, p, qcoeff, dqcoeff, &eob, scan_order,
&quant_param);
#endif // CONFIG_AV1_HIGHBITDEPTH
+
+ if (cpi->oxcf.enable_rate_guide_deltaq) {
+ const int rate_cost = rate_estimator(qcoeff, eob, tx_size);
+ cpi->prep_rate_estimates[(mi_row / mb_step) * cpi->frame_info.mi_cols +
+ (mi_col / mb_step)] = rate_cost;
+ }
+
av1_inverse_transform_block(xd, dqcoeff, 0, DCT_DCT, tx_size, dst_buffer,
dst_buffer_stride, eob, 0);
WeberStats *weber_stats =
@@ -440,6 +479,54 @@
}
}
+static void ext_rate_guided_quantization(AV1_COMP *cpi) {
+ // Calculation uses 8x8.
+ const int mb_step = mi_size_wide[cpi->weber_bsize];
+ // Accumuate to 16x16
+ const int block_step = mi_size_wide[BLOCK_16X16];
+
+ const char *filename = cpi->oxcf.rate_distribution_info;
+ FILE *pfile = fopen(filename, "r");
+ if (pfile == NULL) {
+ printf("Can't open file %s. Default deltaq-mode=3 will be used.\n",
+ filename);
+ fclose(pfile);
+ return;
+ }
+
+ double ext_rate_sum = 0.0;
+ for (int row = 0; row < cpi->frame_info.mi_rows; row += block_step) {
+ for (int col = 0; col < cpi->frame_info.mi_cols; col += block_step) {
+ float val;
+ fscanf(pfile, "%f", &val);
+ ext_rate_sum += val;
+ cpi->ext_rate_distribution[(row / mb_step) * cpi->frame_info.mi_cols +
+ (col / mb_step)] = val;
+ }
+ }
+ fclose(pfile);
+
+ int uniform_rate_sum = 0;
+ for (int row = 0; row < cpi->frame_info.mi_rows; row += block_step) {
+ for (int col = 0; col < cpi->frame_info.mi_cols; col += block_step) {
+ int rate_sum = 0;
+ for (int r = 0; r < block_step; r += mb_step) {
+ for (int c = 0; c < block_step; c += mb_step) {
+ const int mi_row = row + r;
+ const int mi_col = col + c;
+ rate_sum += cpi->prep_rate_estimates[(mi_row / mb_step) *
+ cpi->frame_info.mi_cols +
+ (mi_col / mb_step)];
+ }
+ }
+ uniform_rate_sum += rate_sum;
+ }
+ }
+
+ const double scale = uniform_rate_sum / ext_rate_sum;
+ cpi->ext_rate_scale = scale;
+}
+
void av1_set_mb_wiener_variance(AV1_COMP *cpi) {
AV1_COMMON *const cm = &cpi->common;
const SequenceHeader *const seq_params = cm->seq_params;
@@ -473,6 +560,9 @@
// Determine whether to turn off several intra coding tools.
automatic_intra_tools_off(cpi, sum_rec_distortion, sum_est_rate);
+ // Read external rate distribution and use it to guide delta quantization
+ if (cpi->oxcf.enable_rate_guide_deltaq) ext_rate_guided_quantization(cpi);
+
const BLOCK_SIZE norm_block_size = cm->seq_params->sb_size;
cpi->norm_wiener_variance = estimate_wiener_var_norm(cpi, norm_block_size);
const int norm_step = mi_size_wide[norm_block_size];
@@ -515,8 +605,67 @@
aom_free_frame_buffer(&cm->cur_frame->buf);
}
+static int get_rate_guided_quantizer(AV1_COMP *const cpi, BLOCK_SIZE bsize,
+ int mi_row, int mi_col) {
+ // Calculation uses 8x8.
+ const int mb_step = mi_size_wide[cpi->weber_bsize];
+ // Accumuate to 16x16
+ const int block_step = mi_size_wide[BLOCK_16X16];
+ double sb_rate_hific = 0.0;
+ double sb_rate_uniform = 0.0;
+ for (int row = mi_row; row < mi_row + mi_size_wide[bsize];
+ row += block_step) {
+ for (int col = mi_col; col < mi_col + mi_size_high[bsize];
+ col += block_step) {
+ sb_rate_hific +=
+ cpi->ext_rate_distribution[(row / mb_step) * cpi->frame_info.mi_cols +
+ (col / mb_step)];
+
+ for (int r = 0; r < block_step; r += mb_step) {
+ for (int c = 0; c < block_step; c += mb_step) {
+ const int this_row = row + r;
+ const int this_col = col + c;
+ sb_rate_uniform +=
+ cpi->prep_rate_estimates[(this_row / mb_step) *
+ cpi->frame_info.mi_cols +
+ (this_col / mb_step)];
+ }
+ }
+ }
+ }
+ sb_rate_hific *= cpi->ext_rate_scale;
+
+ const double weight = 1.0;
+ const double rate_diff =
+ weight * (sb_rate_hific - sb_rate_uniform) / sb_rate_uniform;
+ double scale = pow(2, rate_diff);
+
+ scale = scale * scale;
+ double min_max_scale = AOMMAX(1.0, get_max_scale(cpi, bsize, mi_row, mi_col));
+ scale = 1.0 / AOMMIN(1.0 / scale, min_max_scale);
+
+ AV1_COMMON *const cm = &cpi->common;
+ const int base_qindex = cm->quant_params.base_qindex;
+ int offset =
+ av1_get_deltaq_offset(cm->seq_params->bit_depth, base_qindex, scale);
+ const DeltaQInfo *const delta_q_info = &cm->delta_q_info;
+ const int max_offset = delta_q_info->delta_q_res * 10;
+ offset = AOMMIN(offset, max_offset - 1);
+ offset = AOMMAX(offset, -max_offset + 1);
+ int qindex = cm->quant_params.base_qindex + offset;
+ qindex = AOMMIN(qindex, MAXQ);
+ qindex = AOMMAX(qindex, MINQ);
+ if (base_qindex > MINQ) qindex = AOMMAX(qindex, MINQ + 1);
+
+ return qindex;
+}
+
int av1_get_sbq_perceptual_ai(AV1_COMP *const cpi, BLOCK_SIZE bsize, int mi_row,
int mi_col) {
+ if (cpi->oxcf.enable_rate_guide_deltaq) {
+ return get_rate_guided_quantizer(cpi, bsize, mi_row, mi_col);
+ }
+
AV1_COMMON *const cm = &cpi->common;
const int base_qindex = cm->quant_params.base_qindex;
int sb_wiener_var = get_var_perceptual_ai(cpi, bsize, mi_row, mi_col);
diff --git a/av1/encoder/encoder.h b/av1/encoder/encoder.h
index 6852539..a95ea25 100644
--- a/av1/encoder/encoder.h
+++ b/av1/encoder/encoder.h
@@ -1071,6 +1071,15 @@
// CONFIG_PARTITION_SEARCH_ORDER.
const char *partition_info_path;
+ // The flag that indicates whether we use an external rate distribution to
+ // guide adaptive quantization. It requires --deltaq-mode=3. The rate
+ // distribution map file name is stored in |rate_distribution_info|.
+ unsigned int enable_rate_guide_deltaq;
+
+ // The input file of rate distribution information used in all intra mode
+ // to determine delta quantization.
+ const char *rate_distribution_info;
+
// Exit the encoder when it fails to encode to a given level.
int strict_level_conformance;
@@ -3368,6 +3377,23 @@
WeberStats *mb_weber_stats;
/*!
+ * Buffer to store rate cost estimates for each macro block (8x8) in the
+ * preprocessing stage used in allintra mode.
+ */
+ int *prep_rate_estimates;
+
+ /*!
+ * Buffer to store rate cost estimates for each 16x16 block read
+ * from an external file, used in allintra mode.
+ */
+ double *ext_rate_distribution;
+
+ /*!
+ * The scale that equals sum_rate_uniform_quantizer / sum_ext_rate.
+ */
+ double ext_rate_scale;
+
+ /*!
* Buffer to store MB variance after Wiener filter.
*/
BLOCK_SIZE weber_bsize;
diff --git a/av1/encoder/encoder_alloc.h b/av1/encoder/encoder_alloc.h
index 6b0f661..a4aef85 100644
--- a/av1/encoder/encoder_alloc.h
+++ b/av1/encoder/encoder_alloc.h
@@ -311,6 +311,14 @@
aom_free(cpi->mb_weber_stats);
cpi->mb_weber_stats = NULL;
+ if (cpi->oxcf.enable_rate_guide_deltaq) {
+ aom_free(cpi->prep_rate_estimates);
+ cpi->prep_rate_estimates = NULL;
+
+ aom_free(cpi->ext_rate_distribution);
+ cpi->ext_rate_distribution = NULL;
+ }
+
aom_free(cpi->mb_delta_q);
cpi->mb_delta_q = NULL;
}
